Learning color constancy

Brian Funt, Vlad Cardei, Jacobus J Barnard

Research output: Chapter in Book/Report/Conference proceedingConference contribution

102 Citations (Scopus)

Abstract

We decided to test a surprisingly simple hypothesis; namely, that the relationship between an image of a scene and the chromaticity of scene illumination could be learned by a neural network. The thought was that if this relationship could be extracted by a neural network, then the trained network would be able to determine a scene's Illuminant from its image, which would then allow correction of the image colors to those relative to a standard illuminance, thereby providing color constancy. Using a database of surface reflectances and illuminants, along with the spectral sensitivity functions of our camera, we generated thousands of images of randomly selected illuminants lighting 'scenes' of 1 to 60 randomly selected reflectances. During the learning phase the network is provided the image data along with the chromaticity of its illuminant. After training, the network outputs (very quickly) the chro-maticity of the illumination given only the image data. We obtained surprisingly good estimates of the ambient illumination lighting from the network even when applied to scenes in our lab that were completely unrelated to the training data.

Original languageEnglish (US)
Title of host publicationFinal Program and Proceedings - IS and T/SID Color Imaging Conference
Pages58-60
Number of pages3
StatePublished - 1996
Externally publishedYes
EventFinal Program and Proceedings of the 4th IS and T/SID Color Imaging Conference: Color Science, Systems and Applications - Scottsdale, AZ, United States
Duration: Nov 19 1996Nov 22 1996

Other

OtherFinal Program and Proceedings of the 4th IS and T/SID Color Imaging Conference: Color Science, Systems and Applications
CountryUnited States
CityScottsdale, AZ
Period11/19/9611/22/96

Fingerprint

Lighting
Color
Neural networks
Cameras

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Funt, B., Cardei, V., & Barnard, J. J. (1996). Learning color constancy. In Final Program and Proceedings - IS and T/SID Color Imaging Conference (pp. 58-60)

Learning color constancy. / Funt, Brian; Cardei, Vlad; Barnard, Jacobus J.

Final Program and Proceedings - IS and T/SID Color Imaging Conference. 1996. p. 58-60.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Funt, B, Cardei, V & Barnard, JJ 1996, Learning color constancy. in Final Program and Proceedings - IS and T/SID Color Imaging Conference. pp. 58-60, Final Program and Proceedings of the 4th IS and T/SID Color Imaging Conference: Color Science, Systems and Applications, Scottsdale, AZ, United States, 11/19/96.
Funt B, Cardei V, Barnard JJ. Learning color constancy. In Final Program and Proceedings - IS and T/SID Color Imaging Conference. 1996. p. 58-60
Funt, Brian ; Cardei, Vlad ; Barnard, Jacobus J. / Learning color constancy. Final Program and Proceedings - IS and T/SID Color Imaging Conference. 1996. pp. 58-60
@inproceedings{0d4b5895343a415ba4d4e5717ace3188,
title = "Learning color constancy",
abstract = "We decided to test a surprisingly simple hypothesis; namely, that the relationship between an image of a scene and the chromaticity of scene illumination could be learned by a neural network. The thought was that if this relationship could be extracted by a neural network, then the trained network would be able to determine a scene's Illuminant from its image, which would then allow correction of the image colors to those relative to a standard illuminance, thereby providing color constancy. Using a database of surface reflectances and illuminants, along with the spectral sensitivity functions of our camera, we generated thousands of images of randomly selected illuminants lighting 'scenes' of 1 to 60 randomly selected reflectances. During the learning phase the network is provided the image data along with the chromaticity of its illuminant. After training, the network outputs (very quickly) the chro-maticity of the illumination given only the image data. We obtained surprisingly good estimates of the ambient illumination lighting from the network even when applied to scenes in our lab that were completely unrelated to the training data.",
author = "Brian Funt and Vlad Cardei and Barnard, {Jacobus J}",
year = "1996",
language = "English (US)",
pages = "58--60",
booktitle = "Final Program and Proceedings - IS and T/SID Color Imaging Conference",

}

TY - GEN

T1 - Learning color constancy

AU - Funt, Brian

AU - Cardei, Vlad

AU - Barnard, Jacobus J

PY - 1996

Y1 - 1996

N2 - We decided to test a surprisingly simple hypothesis; namely, that the relationship between an image of a scene and the chromaticity of scene illumination could be learned by a neural network. The thought was that if this relationship could be extracted by a neural network, then the trained network would be able to determine a scene's Illuminant from its image, which would then allow correction of the image colors to those relative to a standard illuminance, thereby providing color constancy. Using a database of surface reflectances and illuminants, along with the spectral sensitivity functions of our camera, we generated thousands of images of randomly selected illuminants lighting 'scenes' of 1 to 60 randomly selected reflectances. During the learning phase the network is provided the image data along with the chromaticity of its illuminant. After training, the network outputs (very quickly) the chro-maticity of the illumination given only the image data. We obtained surprisingly good estimates of the ambient illumination lighting from the network even when applied to scenes in our lab that were completely unrelated to the training data.

AB - We decided to test a surprisingly simple hypothesis; namely, that the relationship between an image of a scene and the chromaticity of scene illumination could be learned by a neural network. The thought was that if this relationship could be extracted by a neural network, then the trained network would be able to determine a scene's Illuminant from its image, which would then allow correction of the image colors to those relative to a standard illuminance, thereby providing color constancy. Using a database of surface reflectances and illuminants, along with the spectral sensitivity functions of our camera, we generated thousands of images of randomly selected illuminants lighting 'scenes' of 1 to 60 randomly selected reflectances. During the learning phase the network is provided the image data along with the chromaticity of its illuminant. After training, the network outputs (very quickly) the chro-maticity of the illumination given only the image data. We obtained surprisingly good estimates of the ambient illumination lighting from the network even when applied to scenes in our lab that were completely unrelated to the training data.

UR - http://www.scopus.com/inward/record.url?scp=33646225238&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33646225238&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:33646225238

SP - 58

EP - 60

BT - Final Program and Proceedings - IS and T/SID Color Imaging Conference

ER -